Detection method and detection device

Abstract

To provide a method capable of estimating the position of a deposit in addition to detection of the deposit onto at least one of an inner wall and an outer wall of piping.SOLUTION: A detection method in which an arithmetic circuit accessible to a storage device storing a learned model detects the position of a deposit on at least one of an inner wall and an outer wall of a piping of an object to be detected, where the learned model learns relation between the resonance frequency and a mode vector of a piping having a deposit on at least one of the inner wall and outer wall, and the position of the deposit, the resonance frequency and the mode vector of the piping of the object to be detected is input to the learned model, and the position of the deposit on at least one of the inner wall and outer wall of the piping of the object to be detected is acquired from the learned model.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a detection method and a detection device for detecting the position of an attachment on at least one of an inner wall and an outer wall of a pipe. [Background technology]

[0002] Patent Documents 1 and 2 disclose a method for diagnosing the presence or absence of a blockage in a pipe by utilizing vibrations obtained from a vibration sensor. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 6-258296 [Patent Document 2] JP 2014-190753 A Summary of the Invention [Problem to be solved by the invention]

[0004] The present disclosure aims to provide a method for detecting the location of a deposit on at least one of an inner wall and an outer wall of a pipe. [Means for solving the problem]

[0005] The present disclosure includes the following aspects. [1] A detection method in which an arithmetic circuit capable of accessing a storage device that stores a trained model detects the position of an attachment on at least one of an inner wall and an outer wall of a pipe to be detected, The trained model learns the relationship between a resonant frequency and a mode vector of a pipe having a deposit on at least one of an inner wall and an outer wall, and a position of the deposit; A detection method comprising: inputting a resonant frequency and a mode vector of a pipe to be detected into the trained model; and obtaining from the trained model the position of an adhesion on at least one of an inner wall and an outer wall of the pipe to be detected.

[0006] [2] Before inputting the trained model, acquiring a vibration force and a vibration acceleration applied to the pipe to be detected by the vibration; The detection method described in [1] above, further comprising: obtaining the resonant frequency and the mode vector of the piping to be detected from the excitation force and the vibration acceleration.

[0007] [3] The detection method described in [1] or [2] above, wherein the trained model learns the relationship between the resonant frequency and mode vector of the pipe obtained from the excitation force and vibration acceleration applied to the pipe, generated by a simulation model simulating the pipe to be detected, and the position of the adhesion.

[0008] [4] A detection method as described in [1] or [2] above, wherein the trained model has learned the relationship between the resonant frequency and mode vector of a model pipe corresponding to the pipe to be detected, which are obtained from the excitation force and vibration acceleration applied to the model pipe, and the position of the adhesion.

[0009] [5] The trained model learns the relationship between the resonant frequency and the mode vector and the position and mass of the attachment; The detection method according to any one of the above [1] to [4], wherein the trained model acquires the position of the object as well as the mass of the object.

[0010] [6] The detection method according to any one of the above [1] to [5], wherein the resonant frequency and the mode vector are obtained by performing a modal analysis of the excitation force and the vibration acceleration.

[0011] [7] The detection method according to any one of [1] to [6] above, wherein the trained model is trained using a random forest.

[0012] [8] One or more vibration sensors are attached to the pipe to be detected; The detection method according to any one of the above [1] to [7], wherein the one or more vibration sensors acquire the vibration acceleration.

[0013] [9] The vibration sensors are attached to the pipe to be detected in a number of units per meter or more and not more than 5 units per meter. The detection method according to any one of the above [1] to [7], wherein the vibration sensor acquires the vibration acceleration.

[0014]

[10] A detection device for detecting the position of an attachment on at least one of an inner wall and an outer wall of a pipe to be detected, comprising: A storage device that stores the trained model; A calculation circuit, The trained model learns the relationship between a resonant frequency and a mode vector of a pipe having a deposit on at least one of an inner wall and an outer wall, and a position of the deposit; The arithmetic circuit includes: a position information acquisition process for inputting a resonant frequency and a mode vector of the pipe to be detected into the trained model and acquiring a position of an attachment on at least one of an inner wall and an outer wall of the pipe to be detected from the trained model; A detection device that performs the above.

[0015]

[11] The arithmetic circuit further performs, before the position information acquisition process, a vibration data acquisition process for acquiring a vibration force and a vibration acceleration applied to the pipe to be detected by the vibration; A piping data acquisition process for acquiring a resonance frequency and a mode vector of the piping to be detected from the acquired excitation force and the acquired vibration acceleration; The detection device according to

[10] above, Effect of the Invention

[0016] According to the present disclosure, a method is provided that can detect deposits on at least one of the inner and outer walls of a pipe, as well as estimate the location of the deposits. [Brief description of the drawings]

[0017] [Figure 1] 1 is a schematic perspective view showing an example of an application scene of a clogging detection device according to an embodiment of the present disclosure. [Diagram 2] FIG. 2 is a block diagram illustrating an example of a hardware configuration of a detection device according to an embodiment of the present disclosure. [Diagram 3] 1 is a flowchart illustrating an example of a detection method according to an embodiment of the present disclosure. [Figure 4] FIG. 3 is a block diagram illustrating an example of a learning system for generating the trained model shown in FIG. [Diagram 5] 13 is a flowchart illustrating an example of a learning method. [Figure 6A] FIG. 2 is a schematic diagram showing an example (pattern 1) of a simulated piping used to generate a simulation model. [Figure 6B] FIG. 13 is a schematic diagram showing an example (pattern 2) of a simulated piping used to generate a simulation model. [Figure 6C] FIG. 13 is a schematic diagram showing an example (pattern 3) of a simulated piping used to generate a simulation model. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0018] The detection method according to the present disclosure includes: A detection method in which an arithmetic circuit capable of accessing a storage device that stores a trained model detects a position of an attachment on at least one of an inner wall and an outer wall of a pipe to be detected, The trained model learns the relationship between a resonant frequency and a mode vector of a pipe having a deposit on at least one of an inner wall and an outer wall, and a position of the deposit; The method includes inputting the resonant frequency and mode vector of the pipe to be detected into the trained model, and obtaining the position of an adhesion on at least one of the inner wall and outer wall of the pipe to be detected from the trained model.

[0019] The detection device according to the present disclosure comprises: A detection device that detects the position of an attachment on at least one of an inner wall and an outer wall of a pipe to be detected, A storage device that stores the trained model; A calculation circuit, The trained model learns the relationship between a resonant frequency and a mode vector of a pipe having a deposit on at least one of an inner wall and an outer wall, and a position of the deposit; The arithmetic circuit includes: a position information acquisition process for inputting a resonant frequency and a mode vector of the pipe to be detected into the trained model and acquiring a position of an attachment on at least one of an inner wall and an outer wall of the pipe to be detected from the trained model; Execute.

[0020] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. In the following embodiments, the same or similar components are denoted by the same reference numerals.

[0021] 1. Application Examples First, an example of a scene in which a clogging detection device 100 according to an embodiment is applied will be described with reference to Fig. 1. Fig. 1 is a schematic perspective view showing an example of a scene in which the clogging detection device 100 is applied. The clogging detection device 100 is an example of the "detection device" of the present disclosure.

[0022] The clogging detection device 100 detects the position of the clogging in the pipe 3. The pipe 3 to be detected is formed by connecting a plurality of parts with a flange 31, for example. An obstruction 32 is attached to the inner wall of the pipe 3, and blocks the flow of the pipe 3. A vibration sensor 2 is attached to the pipe 3. The vibration sensor 2 measures the vibration acceleration when the pipe 3 is vibrated by an impact hammer 4. The impact hammer 4 measures the vibration force. The clogging detection device 100 includes a calculation circuit (not shown), and the calculation circuit detects the position of the obstruction 32 from the above-mentioned vibration force and vibration acceleration obtained from the vibration sensor 2 and the impact hammer 4, respectively. In FIG. 1, for convenience, a part of the pipe 3 is cut out to show the obstruction 32 inside, and the obstruction 32 is hatched.

[0023] In an industrial manufacturing plant, pipes are installed and various substances flow through the pipes. When identifying the location of a blockage, the flanges are usually released from their connections, the pipe is disassembled into multiple parts, and the inside of each part is checked. According to this embodiment, the location of a blockage can be identified without disassembling the pipe.

[0024] The impact hammer 4 is an example of a means for vibrating the pipe 3. The means and method for vibrating the pipe 3 are not limited as long as the vibrating force can be measured. The vibration may be performed manually or automatically.

[0025] The positions and number of the flanges 31, the mounting positions and number of the vibration sensors 2, and the positions at which vibration is applied are not limited to those shown in FIG. 1, and can be set appropriately.

[0026] 2.Details 2-1.Detection device 2-1-1. Example of clogging detection device configuration Fig. 2 is a block diagram showing an example of a hardware configuration of the clogging detection device 100 of Fig. 1. The clogging detection device 100 includes an input device 11, an output device 12, an arithmetic circuit 13, and a storage device 14. The clogging detection device 100 may further include a communication interface (not shown) for enabling a communication connection between the clogging detection device 100 and an external device (not shown). The communication interface includes an interface circuit, and performs communication according to an existing wired communication standard or wireless communication standard. These components of the clogging detection device 100 do not need to be housed in a single housing, and may be realized by multiple computer systems.

[0027] The vibration sensor 2 and the impact hammer 4 are connected to the input device 11 of the clogging detection device 100. The input device 11 is an interface circuit that connects the clogging detection device 100 with external devices such as the vibration sensor 2 in order to input information to the clogging detection device 100. The output device 12 is an interface circuit that connects the clogging detection device 100 with external devices in order to output information from the clogging detection device 100.

[0028] The arithmetic circuit 13 includes, for example, a CPU (Central Processing Unit), a RAM (Random Access Memory), and a ROM (Read Only Memory), and controls the operation of the clogging detection device 100 in response to information processing. Such information processing is realized by the arithmetic circuit 13 executing a program. The arithmetic circuit 13 is realized by one or more dedicated processors. Functions of the components of the arithmetic circuit 13 may be omitted, replaced, or added as appropriate depending on the embodiment. The arithmetic circuit 13 may be composed of various semiconductor integrated circuits such as a CPU, an MPU, a GPU, a microcomputer, a DSP, an FPGA, or an ASIC.

[0029] The storage device 14 is a medium that stores information such as a program by electrical, magnetic, optical, mechanical, or chemical action so that a computer or other device, machine, etc. can read the recorded information such as the program. The storage device 14 is, for example, an auxiliary storage device such as a hard disk drive or a solid state drive, and stores the trained model 141 and the program executed by the arithmetic circuit 13.

[0030] The trained model 141 is generated by having the model learn the relationship between the resonant frequency and mode vector of the pipe and the position and mass of an attachment on at least one of the inner wall and the outer wall of the pipe.

[0031] 2, the trained model 141 is stored in the storage device 14, but the present disclosure is not limited thereto. For example, the trained model 141 does not need to be stored in the storage device 14 at all times, and may be stored in the storage device 14 when necessary for processing by the arithmetic circuit 13.

[0032] 2-1-2. Detection method Fig. 3 is a flowchart showing an example of a detection method. The detection method of Fig. 3 is executed by the arithmetic circuit 13 of the clogging detection device 100 of Fig. 2. The processing procedure described below is merely an example, and the processing procedure and each process may be changed as much as possible.

[0033] (Vibration data acquisition process S1) First, the arithmetic circuit 13 acquires the excitation force from the impact hammer 4 and the vibration acceleration from the vibration sensor 2 (S1). For example, the arithmetic circuit 13 acquires the excitation force measured by the impact hammer 4 and the vibration acceleration measured by the vibration sensor 2 via the input device 11.

[0034] A plurality of vibration sensors 2 may be attached to the pipe 3. The vibration sensors 2 may be attached to the pipe 3 at a density of 0.1 to 5 per meter. According to the detection method of the present embodiment, the number of vibration sensors 2 attached can be reduced compared to the conventional method.

[0035] If the mounting positions and number of the vibration sensors 2 are the same as those used when acquiring the input information used to train the trained model 141, the accuracy of detecting the obstruction 32 is further improved.

[0036] (Pipe data acquisition process S2) Next, the arithmetic circuit 13 obtains the resonance frequency and mode vector of the piping 3 from the obtained excitation force and vibration acceleration by modal analysis (S2).

[0037] The mode vector of the piping 3 is obtained, for example, as a normalized mode matrix. The resonance frequency and the mode vector of the piping 3 are obtained, for example, by performing modal analysis of the excitation force and the vibration acceleration and using a receptance coupling method.

[0038] (Location information acquisition process S3) Next, the arithmetic circuit 13 inputs the acquired resonant frequency and mode vector to the trained model 141, and causes the trained model 141 to detect the position and mass of the obstruction 32 inside the piping 3 (S3). The position information acquisition process S3 includes an inference process of the position and mass of the obstruction 32 by the trained model 141. The position information includes the position of the obstruction 32 and its mass.

[0039] As described later, the trained model 141 trains using training data including, as correct answer information (output information), the presence or absence of an obstruction 32 in each of 14 divided regions set by dividing the piping into 14 regions, and the mass of the obstruction 32. Therefore, the position information detected by the trained model 141 is applied to the piping 3 in which the same divided regions as above (14 divided regions in this embodiment) are set.

[0040] As described above, the detection method executed by the clogging detection device 100 includes the vibration data acquisition process S1, the piping data acquisition process S2, and the position information acquisition process S3. The position information acquisition process S3 causes the trained model 141 to detect the position and mass of the obstruction 32 inside the piping 3 from the resonance frequency and mode vector of the piping 3. This detection method not only detects the presence or absence of the obstruction 32 inside the piping 3, but also makes it possible to estimate the position and mass of the obstruction 32 with high accuracy without the need to disassemble the piping 3.

[0041] 2-2. Learning system 2-2-1. Example of learning system configuration 4 is a block diagram showing an example of a learning system 200 for generating the trained model 141 shown in FIG. 2. The learning system 200 includes an input device 21, an output device 22, an arithmetic circuit 23, and a storage device 24. The learning system 200 may further include a communication interface (not shown) for enabling a communication connection between the learning system 200 and an external device (not shown). The communication interface includes an interface circuit, and performs communication according to an existing wired communication standard or wireless communication standard. These components of the learning system 200 do not need to be housed in a single housing, and may be realized by multiple computer systems.

[0042] The input device 21 is an interface circuit that connects the learning system 200 to an external device for inputting information to the learning system 200. The output device 22 is an interface circuit that connects the learning system 200 to an external device for outputting information from the learning system 200.

[0043] The arithmetic circuit 23 includes a CPU (Central Processing Unit), RAM (Random Access Memory), ROM (Read Only Memory), etc., and controls the operation of the learning system 200 in response to information processing. Such information processing is realized by the arithmetic circuit 23 executing a program. The arithmetic circuit 23 is realized by one or more dedicated processors. Functions of the components of the arithmetic circuit 23 may be omitted, replaced, or added as appropriate depending on the embodiment. The arithmetic circuit 23 may be composed of various semiconductor integrated circuits such as a CPU, MPU, GPU, microcomputer, DSP, FPGA, ASIC, etc.

[0044] The storage device 24 is a medium that accumulates information such as a program by electrical, magnetic, optical, mechanical, or chemical action so that a computer or other device, machine, etc. can read the recorded information such as the program. The storage device 24 is, for example, an auxiliary storage device such as a hard disk drive or a solid state drive, and stores a model 241, vibration data 242, piping data 243, teacher data 245, a program executed by the arithmetic circuit 23, etc. Details of the model 241, vibration data 242, piping data 243, data set 244, and teacher data 245 will be described later.

[0045] 4, these data 242 to 245 are stored in the storage device 24, but the present disclosure is not limited to this. For example, these data do not need to be stored in the storage device 24 at all times, and may be stored in the storage device 24 when necessary for processing by the arithmetic circuit 23.

[0046] 2-2-2. Example of learning system operation Next, as a learning method, a method for generating the trained model 141 will be described. FIG. 5 is a flowchart showing an example of the learning method. The learning method is executed, for example, by the learning system 200 shown in FIG. 4. The arithmetic circuit 23 of the learning system 200 trains the model 241 to generate the trained model 141. The trained model 141 is generated by supervised learning. The processing procedure described below is merely an example, and the processing procedure and each process may be changed as much as possible.

[0047] (Vibration data preparation process S201) First, the arithmetic circuit 23 prepares the vibration data 242 (S201). For example, the arithmetic circuit 23 acquires the vibration data 242 stored in the storage device 24. The arithmetic circuit 23 may acquire the vibration data 242 from an external device such as a vibration sensor via the input device 21. The vibration data 242 includes the vibration force and the vibration acceleration generated by a simulation model that simulates the piping to be detected.

[0048] 6A is a schematic plan view showing an example (Pattern 1) of the simulated pipe 330 used to generate the simulation model. In the simulation model, the simulated pipe 330 is made of steel, has a length of 2.7 m and a diameter of 114 mm, and both ends are free ends. The simulated pipe 330 is divided into 15 parts (No. 1 to No. 15) in the longitudinal direction, and a 1000 kg / m2 density pipe is applied to one of the divided parts (No. 4 in FIG. 6A) as an attachment. 3 The simulated obstruction 332 is placed in a predetermined volume.

[0049] A total of 15 vibration sensors (not shown) are attached, one at the center of each divided area in the longitudinal direction of the simulated pipe 330. Divided area No. 15 is vibrated with an impact hammer (not shown), and the vibration force is obtained from the impact hammer, and the vibration acceleration of each divided area is obtained from each vibration sensor. In this way, 15 vibration accelerations are obtained from one pattern.

[0050] The divided regions to be vibrated are not particularly limited. However, the vibration position is the same between multiple data sets described later. The mounting positions and the number of vibration sensors 2 and the number of divided regions are not limited to those in this embodiment and can be set appropriately.

[0051] (Piping data preparation process S202) Next, the arithmetic circuit 23 prepares piping data 243 for learning (S202). The piping data 243 includes the resonance frequency and the mode vector of the simulated piping 330 acquired from the vibration data 242.

[0052] The mode vector is obtained, for example, as a normalized mode matrix. The above-mentioned resonance frequency and mode vector are obtained, for example, by performing modal analysis on the vibration data 242 and using a receptance coupling method.

[0053] A total of 16 components are obtained from the vibration data 242, including one component related to the resonance frequency and 15 components related to the mode vector obtained from the 15 vibration accelerations. The 16 components are treated as one unit, and five units in the first to fifth modes are created. A total of 80 components are included in the five units. In other words, the piping data 243 includes a total of 80 components per pattern as the resonance frequency and the mode vector. The number of components included in the piping data 243 is not particularly limited. The number of components included in the piping data 243 is set appropriately.

[0054] (Data set preparation process S203) Next, the arithmetic circuit 23 prepares a data set 244 (S203). The data set 244 includes the piping data 243 as input information and the mass of the simulated obstruction object 332 in each divided area as answer information (output information). The mass of the simulated obstruction object 332 in each divided area indicates the divided area number in which the simulated obstruction object 332 is located and the mass of the simulated obstruction object 332.

[0055] The correct answer information corresponds to pattern 1, and the mass of a predetermined simulated obstruction object 332 is set in divided area No. 4, and the mass is set to zero in each of the other divided areas. The data set 244 includes the piping data 243 (the above 80 components) and the position and mass of the simulated obstruction object 332 associated therewith.

[0056] (Repeat processing) Another simulation model is generated by preparing 44 more patterns of simulated piping 330 (patterns) by changing the position, number, and density (mass) of the simulated obstruction 332. For the other simulation models, the above-mentioned vibration data preparation process S201, piping data preparation process S202, and data set preparation process S203 are performed to prepare 44 new data sets 244.

[0057] The number of times the process is repeated is not particularly limited. The vibration data preparation process S201, the piping data preparation process S202, and the data set preparation process S203 may be repeated to obtain a desired number of data sets 244. In this embodiment, a total of 45 patterns of data sets 244 are prepared.

[0058] 6B is a schematic plan view showing an example (pattern 2) of the simulated piping 330 used to generate the simulation model. In pattern 2, the density of 1000 kg / m 3 The simulated obstructions 332 are arranged in the respective positions.

[0059] 6C is a schematic plan view showing an example (pattern 3) of the simulated piping 330 used to generate the simulation model. In pattern 3, the density of the simulated piping is 600 kg / m 3 A simulated obstruction 332 is placed.

[0060] The positions, number and density of the simulated obstructions 332 used to generate the simulation model are not limited to those in this embodiment, and can be set appropriately.

[0061] (Teacher data preparation process S204) Next, the arithmetic circuit 23 prepares 45 patterns of data sets 244 as teacher data 245 (S204). The teacher data 245 includes at least one data set 244, and preferably two or more data sets 244.

[0062] (Learning process S205) After the teacher data preparation process S204, the arithmetic circuit 23 executes machine learning for the model 241 using the teacher data 245 (S205). This learning process S205 includes having the model 251 learn the correspondence results between the piping data 243 and the positions and masses of the simulated obstruction objects 332 in each divided region by a random forest, thereby generating the trained model 141.

[0063] The model 251 is, for example, a decision tree model that performs machine learning using a decision tree method including a classification tree. The model 251 may be a learning model having a support vector machine or a neural network structure, for example, a convolutional neural network (CNN).

[0064] The learning may be performed, for example, by a boosting learning algorithm such as AdaBoost, logistic regression, decision jungle, or k-nearest neighbors.

[0065] As described above, the learning method is executed by the learning system 200. The learning system 200 includes a vibration data preparation process S201, a piping data preparation process S202, a data set preparation process S203, a teacher data preparation process S204, and a learning process S205. The vibration data 242 and the piping data 243 are acquired for each divided region. By using such a learning method, a trained model 141 that can improve the accuracy of estimating the position and mass of the obstruction 32 can be obtained.

[0066] (Modification) Although the embodiment of the present disclosure has been described in detail above, the above description is merely an example of the present disclosure in every respect. Various improvements and modifications can be made without departing from the scope of the present disclosure. For example, the following modifications are possible. In the following, the same reference numerals are used for the same components as in the above embodiment, and the description of the same points as in the above embodiment will be omitted as appropriate. The following modifications can be combined as appropriate.

[0067] In the present embodiment, the clogging detection device 100 and the learning system 200 are realized by different computer systems, but this is not limiting. The clogging detection device 100 and the learning system 200 may be realized by a single computer system.

[0068] In this embodiment, a method for detecting the position and mass of a blockage inside the pipe 3 has been described, but is not limited thereto. According to the detection device and detection method of the present disclosure, it is possible to detect the position and mass of a deposit that has not caused a blockage and that has adhered to the inner wall of the pipe, as well as the deposit that has adhered to the outer wall.

[0069] In this embodiment, the simulated obstruction 332 is used as an attachment when generating the trained model 141, but is not limited to this. Instead of the simulated obstruction 332, an attachment that does not lead to obstruction and that adheres to the inner wall of the pipe, or an attachment that adheres to the outer wall may be used.

[0070] In the present embodiment, a method for detecting the position and mass of an object adhering to the pipe 3 has been described, but the present disclosure is not limited thereto. The detection device and detection method according to the present disclosure can at least detect the position of an object adhering to the pipe 3.

[0071] In this embodiment, the excitation force is obtained from the impact hammer 4, but is not limited to this. The excitation force may be obtained from the vibration sensor 2. In this case, the vibration sensor 2 is connected to the input device 11. In this case, instead of the impact hammer 4, excitation may be performed using a means that cannot measure the excitation force, such as a wooden mallet or a plastic hammer.

[0072] In this embodiment, the trained model 141 is generated using the excitation force and vibration acceleration generated by a simulation model simulating the piping 3 to be detected, but this is not limiting. The trained model 141 may be generated using the excitation force and vibration acceleration acquired by vibrating a model piping corresponding to the piping 3 to be the actual detection target. The trained model 141 may also be generated using the excitation force and vibration acceleration already acquired for the piping 3 to be the actual detection target.

[0073] In this embodiment, the vibration acceleration of the pipe 3 caused by the excitation is acquired, but the invention is not limited to this. Instead of the vibration acceleration, the displacement amount or velocity of the amplitude caused by the excitation may be acquired.

[0074] In the present embodiment, the excitation force is acquired, but the present invention is not limited to this. The resonance frequency and the mode vector of the pipe may be acquired from the vibration acceleration and other information acquired from the vibration sensor 2. [Industrial Applicability]

[0075] The detection method and detection device disclosed herein can estimate the location of adhesions inside and outside of a pipe with high accuracy, and are therefore suitable for use in, for example, the maintenance of pipes used in factories. [Explanation of symbols]

[0076] 100 Clogging detection device 11 Input Devices 12 Output Devices 13 Arithmetic circuit 14 Storage device 141 trained models 2. Vibration Sensor 3 Piping 31 Flange 32 Obstruction 4. Impact Hammer 200 Learning System 21 Input Devices 22 Output Devices 23 Arithmetic circuit 24 Storage device 241 Model 242 Vibration Data 243 Piping Data 244 Dataset 245 Teacher Data 330 Simulated piping 332 Simulated obstruction

Claims

1. A detection method in which an arithmetic circuit capable of accessing a storage device that stores a trained model detects a position of an attachment on at least one of an inner wall and an outer wall of a pipe to be detected, The trained model learns the relationship between a resonant frequency and a mode vector of a pipe having a deposit on at least one of an inner wall and an outer wall, and a position of the deposit; inputting a resonant frequency and a mode vector of a pipe to be detected into the trained model, and obtaining a position of an attachment on at least one of an inner wall and an outer wall of the pipe to be detected from the trained model; A detection method comprising:

2. Prior to inputting the trained model, acquiring a vibration force and a vibration acceleration applied to the pipe to be detected by the vibration; The detection method according to claim 1 , further comprising: acquiring the resonant frequency and the mode vector of the pipe to be detected from the excitation force and the vibration acceleration.

3. The detection method according to claim 1 or 2, wherein the trained model learns the relationship between the resonant frequency and mode vector of the pipe obtained from the excitation force and vibration acceleration applied to the pipe, which are generated by a simulation model simulating the pipe to be detected, and the position of the attachment.

4. The detection method described in claim 1 or 2, wherein the trained model learns the relationship between the resonant frequency and mode vector of a model piping corresponding to the piping to be detected, which are obtained from the excitation force and vibration acceleration applied to the model piping, and the position of the adhesion.

5. The trained model learns the relationship between the resonant frequency and the mode vector and the position and mass of the attachment, The detection method according to claim 1 or 2, wherein the trained model acquires the mass of the object along with the position of the object.

6. The detection method according to claim 1 , wherein the resonant frequency and the mode vector are obtained by performing a modal analysis of the excitation force and the vibration acceleration.

7. The detection method according to claim 1 or 2, wherein the trained model is trained using a random forest.

8. One or more vibration sensors are attached to the pipe to be detected, The method according to claim 1 or 2, wherein the one or more vibration sensors acquire the vibration acceleration.

9. The vibration sensors are attached to the pipe to be detected in an amount of 0.1 units / m or more and 5 units / m or less, The detection method according to claim 1 , wherein the vibration sensor acquires the vibration acceleration.

10. A detection device that detects the position of an attachment on at least one of an inner wall and an outer wall of a pipe to be detected, A storage device that stores the trained model; A calculation circuit, The trained model learns the relationship between a resonant frequency and a mode vector of a pipe having a deposit on at least one of an inner wall and an outer wall, and a position of the deposit; The arithmetic circuit includes: a position information acquisition process for inputting a resonant frequency and a mode vector of the pipe to be detected into the trained model and acquiring, from the trained model, a position of an attachment on at least one of an inner wall and an outer wall of the pipe to be detected; A detection device that performs the above.

11. The arithmetic circuit further performs the following steps before the position information acquisition process: a vibration data acquisition process for acquiring a vibration force and a vibration acceleration applied to the pipe to be detected by the vibration; A piping data acquisition process for acquiring a resonance frequency and a mode vector of the piping to be detected from the acquired excitation force and the acquired vibration acceleration; The detection device of claim 10 , further comprising:

Images